Personalized language offline learning method based on deep neural network
A technology of deep neural network and neural network, applied in the field of offline learning of personalized language
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Publication Date
- 2021-03-26
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
technical field
[0001] The invention belongs to the technical field of speech recognition, and in particular relates to a method for offline learning of personalized language based on a deep neural network. Background technique
[0002] With the continuous development of the field of artificial intelligence speech recognition, more and more attention has been paid to the problem of offline personality language recognition. The current traditional offline speech recognition technology only conducts deep neural network training and recognition for specific languages, and conducts offline device speech recognition through acoustic models and language models. Usually, manufacturers will train one or more languages with a deep neural network to obtain an acoustic model, and then define different language models for different products. The acoustic model defines the language, and the language model defines the recognition content. After the offline voice recognition device ente...
Examples
Embodiment Construction
[0029] Specific embodiments of the present invention will be further described in detail below.
[0030] The offline personalized language self-learning method based on the deep neural network of the present invention, such as figure 1 shown, including the following steps:
[0031] S11. The system enters the learning state;
[0032] S12. The user uses the language to be learned by the system to read aloud, and the system continues to collect voice; perform deep neural network operations on the collected voice data to obtain neural network acoustic features, and store the features;
[0033] The voice reading performed in step 2 is generally not shorter than 2 seconds.
[0034] S13. Step S12 is repeated multiple times, and the system performs deep neural network operations on the voice data collected each time to obtain the neural network acoustic features;
[0035] For the multiple neural network acoustic features obtained, the marginal distance is calculated in pairs;
[0...